1 Phosphorus Recovery from Microbial Biofuel Residual Using Microwave Peroxide 2 Digestion and Anion Exchange 3 4 McKay Gifforda*, Jianyong Liub, Bruce E. Rittmannc, Raveender Vannelac, Paul 5 Westerhoffa 6 7 * 8 a 9 Tempe, Box 5306, AZ 85287-5306; phone: 480-965-2885; fax: 480-965-0557; email: 10 Corresponding author: Arizona State University, School of Sustainable Engineering and The Built Environment, mac.gifford@asu.edu 11 12 Affiliations 13 a 14 Tempe, AZ 85287-5306 15 b 16 Road, Shanghai 200444, P. R. China 17 c 18 Institute, Tempe, AZ 85287-5701 Arizona State University, School of Sustainable Engineering and the Built Environment, School of Environmental and Chemical Engineering, Shanghai University, 333 Nanchen Arizona State University, Swette Center for Environmental Biotechnology, Biodesign 19 20 Last revision: November 10, 2014 21 In preparation for: Water Research (Elsevier) 22 1 23 24 Abstract Sustainable production of microalgae for biofuel requires efficient phosphorus (P) 25 utilization, which is a limited resource and vital for global food security. This research 26 tracks the fate of P through biofuel production and investigates P recovery from the 27 biomass using the cyanobacterium Synechocystis sp. PCC 6803. Our results show that 28 Synechocystis contained 1.4% P dry weight. After crude lipids were extracted (e.g., for 29 biofuel processing), 92% of the intracellular P remained in the residual biomass, indicating 30 phospholipids comprised only a small percentage of cellular P. We estimate a majority of 31 the P is primarily associated with nucleic acids. Advanced oxidation using hydrogen 32 peroxide and microwave heating released 92% of the cellular P into orthophosphate. We 33 then recovered the orthophosphate from the digestion matrix using two different types of 34 anion exchange resins. One resin impregnated with iron nanoparticles adsorbed 98% of 35 the influent P through 20 bed volumes, but only released 23% during regeneration. A 36 strong-base anion exchange resin adsorbed 87% of the influent P through 20 bed volumes 37 and released 50% of it upon regeneration. This recovered P subsequently supported 38 growth of Synechocystis. This proof-of-concept recovery process reduced P demand of 39 biofuel microalgae by 54%. 40 41 Keywords 42 Microbial Biofuel, Phosphorus Recovery, Oxidation, Anion Exchange, Iron Nanoparticles 43 2 44 1. Introduction 45 There is an urgent need to find energy replacements for fossil fuels, whose 46 combustion releases known and suspected human carcinogens and greenhouse gases into 47 the atmosphere. One promising alternative is biofuel, which provides renewable energy 48 with net greenhouse gas emissions significantly lower than fossil fuel (Batan et al. 2010). 49 Biofuel derived from microalgae offers several advantages over biofuel from terrestrial 50 plants: it does not compete with food crops for arable land, it can be continuously 51 harvested, and it provides a much higher areal yield (Rittmann 2008, Schenk et al. 2008). 52 Microalgae biofuel production requires several inputs, including water, sunlight, 53 carbon dioxide, and nutrients – particularly nitrogen (N) and phosphorus (P). During lipid 54 extraction from microalgae biomass for liquid fuels, most of the N and P are discarded, 55 requiring new nutrients for subsequent growth. Should microalgae become a significant 56 replacement for fossil fuel in the future, the requirements for biomass growth would create 57 a huge nutrient demand, rivaling that of agriculture (Erisman et al. 2010). Thus, capturing 58 and recycling nutrients represents a significant opportunity for making large-scale 59 cultivation of microalgae more sustainable (Clarens et al. 2010). 60 Nutrient recycling is particularly essential for P. Unlike N, which can be fixed 61 from the atmosphere through the Haber-Bosch method (Huo et al. 2012), P is mined from 62 ore that has finite stocks. World reserves of accessible P are estimated as 65,000 million Abbreviations: ATP, adenosine triphosphate, DI, deionized water; EBCT, empty bed contact time; FAME, fatty acid methyl esters; HAX, hybrid anion exchange; ortho-PO43-, orthophosphate; P, phosphorus; PG, phosphatidylglycerol; SBAX, strong base anion exchange. 3 63 metric tons (USGS 2011), and these are non-renewable and not substitutable. Depletion 64 of economically affordable P may bring about international crises due to the essential role 65 of P fertilizer for global food production (Cordell et al. 2009). Farmers in developing 66 countries could be disproportionately harmed (Childers et al. 2011). Sustainable microbial 67 biofuel production demands efficient nutrient recycling to prevent biofuel from becoming 68 an enormous P demand competing with food production. 69 This research develops a proof-of-concept process for P-recovery from microalgae 70 after extraction of lipids. The research objective is to track P through biofuel production 71 and then recover P from residual biomass in a reusable form by using advanced oxidation 72 to release the P for efficient ion exchange capture. The reusable form provides 73 bioavailable P that supports microalgae growth. 74 We selected the cyanobacteria for this work because it is an excellent candidate for 75 future utilization in large-scale biomass cultivation, particularly when energy efficiency in 76 biosynthesis of fatty acids is crucial (Wijffels et al. 2013). Specifically we use 77 Synechocystis sp. PCC 6803, which is a prokaryotic autotroph, Gram negative and able to 78 withstand a wide range of environmental conditions. Lipids in the form of diacylglycerols 79 are available in an extensive network of thylakoid membranes (van de Meene et al. 2006, 80 Vermaas 2001). 81 production such as high lipid content (Vermaas 1996) because the entire genome has been 82 sequenced (Kaneko et al. 1996). It may be genetically manipulated for specific traits favorable for biofuel 83 84 1.1. P Recovery 4 85 To recover P from microbial biomass we first release organic-bound P as inorganic 86 orthophosphate (ortho-PO43-). This is necessary to improve the efficiency of the 87 subsequent capture since ortho-PO43- is more reactive. It also mitigates heterotrophic 88 contamination of the biomass culture, which can occur after long run periods or with 89 accumulation of inactive cells (Mata et al. 2010). Subsequently, we selectively capture the 90 ortho-PO43- from the liquid in a usable form. This is necessary to isolate and purify the 91 ortho-PO43-, allowing accurate and controlled dosing into the aqueous growth media during 92 reuse. It also concentrates the ortho-PO43- solution to minimize handling or hauling. This 93 subsection gives the impetus for the technologies we selected to accomplish those goals. 94 Many P-recovery methods are available (de-Bashan and Bashan 2004, Morse et al. 95 1998, Rittmann et al. 2011). We selected an advanced oxidation process using hydrogen 96 peroxide and microwave heating to release organic P from the residual biomass. Advanced 97 oxidation creates hydroxyl free radicals that are highly effective for attacking organic 98 matter to release ortho-PO43- (Liao et al. 2005). This transformation may involve oxidation 99 and hydrolysis reactions. While it may be possible to find technologies that are less 100 energy-intensive, such as enzymatic hydrolysis or microbial fuel cells (Rittmann et al. 101 2011), or that do not dilute the biomass with additional liquid such as supercritical carbon 102 dioxide (Blocher et al. 2012, Soh and Zimmerman 2011), advanced oxidation demonstrates 103 the principle for releasing PO43-. 104 We capture ortho-PO43- using ion exchange since it recovers a liquid concentrate 105 that is preferable for nutrient reuse during aquatic microalgae production. Other common 106 recovery techniques such as aluminum adsorption or struvite precipitation (de-Bashan and 107 Bashan 2004) produce complex or low solubility solids which may be better suited for 5 108 agricultural application. We evaluated two anion-exchange resins having distinctly 109 different properties. The first was a hybrid anion exchange resin (HAX) impregnated with 110 iron (hydr)oxide nanoparticles (Layne RT, Layne Christensen). It is reported to have a 111 high sorption capacity and selectivity for ortho-PO43- (Sengupta 2013) and the ability to 112 release a high concentration ortho-PO43- solution upon regeneration (Blaney et al. 2007, 113 Midorikawa et al. 2008). The second was a type-1 strong-base anion exchange resin 114 (SBAX) with quaternary amine functional groups in chloride ion form (21K-XLT, 115 Dowex). It has a general anion-exchange capacity of 1.4 equivalents/L. It has previously 116 been used for uranium (Stucker et al. 2011) and chromium (Rees-Nowak et al. 2005) 117 removal, but has yet to be tested for phosphate recovery. 118 While the individual P recovery technologies employed in this study are not novel 119 by themselves, their usage together such that the P completes an entire use and reuse cycle 120 is. It is also the first study we know of to apply these technologies in the context of 121 microbial biofuel production. Thus this study serves as a proof-of-concept that proposes 122 an approach and can inform future optimization. 123 124 125 1.2. Microbial P To focus the recovery efforts properly, this subsection estimates where P in 126 Synechocystis is located based on literature review. Others have done this for several 127 marine microalgae (Geider and La Roche 2002, Sterner and Elser 2002) but not 128 specifically for Synechocystis. Biochemical fractions in cells can vary based on growth 129 conditions (Sheng et al. 2011a) but this provides clues for understanding the fate of P after 130 lipid processing. Figure 1 summarizes the expected location of P in a Synechocystis cell. 6 131 P may be located within adenosine triphosphate (ATP), lipids, and nucleic acid. The 132 following three paragraphs individually analyze them. 133 ATP contains over 18% P by weight (C10H16N5O13P3), but comprises less than 30 134 µg per g of cell mass. P associated with ATP is therefore 5 µg per g of the cell mass, 135 which is a negligible contributor of the total cell P. The diphosphate form ADP and 136 monophosphate form AMP are smaller fractions of the cell mass with less incorporated P 137 and are also negligible contributors of cellular P storage. 138 The P content associated with lipid is a function of the fraction of lipid that is 139 phospholipid and the fraction of phospholipid that is P. The predominant phospholipid 140 head within cyanobacteria is phosphatidylglycerol (PG), which is the only phospholipid 141 associated with thylakoid membranes in Synechocystis sp. PCC 6803 (Hajime and Murata 142 2007). PG has an elemental composition of C8H12O10P. The most prevalent fatty acid 143 chain in Synechocystis is C16:0, or palmitic acid (Sheng et al. 2011b), which has an 144 elemental composition of C16H32O2. Assuming that all phospholipids within Synechocystis 145 are the diacylglycerol PG with two palmitic acid molecules, the overall elemental formula 146 for a phospholipid molecule is C40H76O14P. That means phospholipid is approximately 147 3.8% P by weight. PG-based lipids comprise approximately 14% of all lipids in 148 Synechocystis (Sakurai et al. 2006), and lipids represent approximately 10% of the 149 biological makeup of the overall cell (Shastri and Morgan 2005). Combining these 150 estimates gives the theoretical amount of P associated with lipid in Synechocystis sp. PCC 151 6803 as 0.05% of the total cell weight, or 2% of the total cell P. A genetically altered high 152 lipid strain containing 50% crude lipids could then have as high as 0.3% of the total cell 153 weight be P associated with lipid. For this reason, we do not expect much P in the lipids. 7 154 We estimate the P content associated with DNA and RNA by comparing its 155 biological composition with its elemental composition. Synechocystis sp. PCC 6803 is 156 approximately 3% DNA and 17% RNA by weight (Shastri and Morgan 2005). DNA and 157 RNA are 10% P by weight (Sterner and Elser 2002). Therefore, P associated with DNA 158 comprises 0.3% of the total cell weight, and P associated with RNA is 1.7% of the total 159 cell weight. This is respectively 15% and 83% of the total cellular P. We consequently 160 expect that most of the cellular P will be in nucleic acid. This was also observed in other 161 studies on lake bacteria where P associated with RNA comprised a majority of the total 162 cell P (Elser et al. 2003, Geider and La Roche 2002). 163 164 165 2. Materials and Methods 166 2.1. Strain, Growth Conditions, and Biomass Production 167 We grew Synechocystis sp. PCC 6803 in BG-11 growth media (Rippka et al. 1979) 168 modified to have five times the normal amount of phosphate (added as K2HPO4) (Kim et 169 al. 2010) in a bench-top photobioreactor in semi-continuous growth mode. We separated 170 biomass from the growth medium by means of centrifugation at 1,500 g for 20 min in 50- 171 mL plastic tubes. We resuspended the cell pellet in 1 mM sodium bicarbonate (Sigma- 172 Aldrich) to rinse away residual medium. We repeated centrifuging and rinsing two times 173 before freeze-drying the final pellet (Labconco Freezone 6) for 2 days at 0.013 mbar and - 174 50ºC in order to obtain an accurate starting dry weight (Sheng et al. 2011b). We collected 175 enough biomass to perform all lipid extraction and P recovery experiments at least in 176 duplicate. 8 177 178 2.2. Lipid Extraction and Transesterification 179 We extracted lipids from the freeze-dried biomass using the Folch Method (Folch 180 et al. 1957) using a 2:1 (V:V) mixture of chloroform (Mallinckrodt) and methanol (Fisher 181 Scientific), since it has a high extraction efficiency for Synechocystis lipids (Sheng et al. 182 2011b). We ground a 300-mg (all weights given as dry weight) sample with agate mortar 183 and pestle, suspended it in 60 mL of Folch solvent, and placed it on a shaker table at 175 184 rpm for 2 days. We filtered the suspension with a glass fiber filter (Whatman GF/B) and 185 then a 0.2-µm polytetrafluoroethylene filter (Whatman). The biomass retained on both 186 filters was the primary residual, and the filtrate contained the extracted crude lipid. For 187 samples undergoing transesterification, we evaporated the solvent from the crude lipid 188 under N2 gas to avoid oxidation of lipids. For samples where no further lipid processing 189 was necessary, we evaporated the solvent by heating on hot plate. 190 We transesterified the crude lipid (Sheng et al. 2011b) by adding 1 mL of 191 methanolic hydrochloric acid (Supelco) and heating the mixture in an 85ºC water bath for 192 2 h. After cooling the mixture to room temperature, we added 0.5 mL of deionized (DI) 193 water and 1 mL of hexane, shook the mixture by hand for 30 s, and allowed the phases to 194 separate. We repeated all transesterification steps two additional times, and then pooled all 195 the hexane. The extracted hexane contained the fatty acid methyl esters (FAME), and the 196 remaining water contained the secondary residual. 197 For experiments tracking the fate of P, we analyzed total P for each biomass, 198 primary residual, crude lipid, secondary residual, and transesterified FAME (at least 199 duplicate samples). 9 200 201 2.3. Advanced Oxidation We scraped primary residual from the dried filters and added it to 60 mL (giving 202 203 3.6 gVSS/L) of 30% ultrapure H2O2 solution (JT Baker Ultrex II) diluted 1:10. We let this 204 mixture stand for 1 hr of pre-digestion under fume-hood ventilation. We digested the 205 mixtures in a microwave (CEM MARS XPress) at 400 W by ramping the temperature up 206 to 170ºC over 10 min and then holding at 170ºC for 10 min per method SW846-3015 207 (USEPA 2008). Others have observed the highest fraction of P release by this peroxide 208 dose and microwave heating temperature (Liao et al. 2005, Wong et al. 2006), and future 209 work may explore varying other conditions to optimize P release. We employed high- 210 pressure microwave vessels to avoid breakage that the high rate of gas evolution could 211 cause. We analyzed duplicate samples before and after oxidation for total P and ortho- 212 PO43-. 213 214 215 2.4. Phosphate Separation We did preliminary investigation of the P separation capacity of each of the two 216 anion exchange resins by placing 3.5 g of fresh resin in a 1.5-cm inner diameter glass 217 column, giving a bed depth of 3.0 cm. We supported the resin with glass beads to ensure 218 even flow distribution. We flushed 100 mL of DI water through the column and allowed 219 air bubbles to escape. Then, we pumped a solution of monobasic sodium phosphate 220 (Mallinkrodt ACS grade) in DI water (concentration 80 mgP/L) through the column at 3.2 221 mL/min to give an empty bed contact time (EBCT) of approximately 2 min (loading rate of 222 4.4 mgP/s/g resin). We periodically took effluent samples for P analysis, and continued 10 223 the experiment until the effluent P concentration stabilized near the influent P 224 concentration. We then desorbed the P using a strong regeneration solution at a pump rate 225 of 0.5 mL/min (EBCT of approximately 10 min) until the effluent P concentration 226 stabilized at nearly zero. The strong regeneration solution used for the HAX resin was 0.1 227 N potassium hydroxide (EMD), and for the SBAX resin was 0.1 N sodium chloride (Sigma 228 Aldrich). We later varied influent P concentration, EBCT, P loading rate, influent pH, and 229 elute contact time in order to optimize column operation. We then tested each resin with biomass after advanced oxidation by pumping the 230 231 60 mL of digested sample through 2.0 g of fresh resin having a bed depth of 1.7 cm. The 232 flow rate was 1.4 mL/min, giving an EBCT of approximately 2 min. We collected the 233 effluent and pumped it through the column two more times to ensure complete capture of 234 phosphate onto the resin. We then recovered retained ortho-PO43- by removing the resin 235 from the column and placing it in 33 mL (11 bed volumes) of strong regeneration solution, 236 which was heated on a 95ºC hot plate, shaken for 24 h, and then decanted. Elution and 237 decanting were repeated two times, and the elution solutions were pooled so that the serial 238 batch elution mimicked a continually stirred tank mixer (CSTM) in series (n = 3). We 239 analyzed the total volume of 100 mL (33 bed volumes) for pH, total P, and ortho-PO43-. 240 We obtained the total mass of P sorbed to each resin by summing the difference 241 between the influent concentration and the effluent concentration for each sample 242 multiplied by the volume treated in the time segment (area above the curve times flow 243 rate). 244 245 2.5. Phosphorus Reuse 11 246 As a confirmatory experiment, recovered P solution was used to culture wild-type 247 Synechocystis sp. PCC 6803 cells. We diluted the recovered P solution to P concentration 248 prescribed by standard BG-11, spiked the other nutrients to standard levels, then added 249 additional bicarbonate to compensate for low aeration in small samples. We inoculated 250 plastic tubes containing 20 mL of the growth media with fresh Synechocystis cells in 251 duplicate. We placed these on a shake table under constant light conditions for one week, 252 and regularly monitored optical density by absorbance at 730 nm. 253 254 255 2.6. Phosphorus Analysis We determined ortho-PO43- colorimetrically with a spectrophotometer (HACH 256 DR5000) using the PhosVer 3 Method (HACH), which is equivalent to Standard Methods 257 4500-P.E (Miner 2006). It directs to add reagent powder to 5 mL of sample and give 2 258 min of reaction time, then measure results at 880 nm. 259 We assayed total P by persulfate digestion (Standard Method 4500-P.B.5) (Miner 260 2006) followed by inductively coupled plasma optical emission spectrometry (ICP-OES). 261 To do this we suspended samples in 50 mL DI water plus 1 mL of concentrated sulfuric 262 acid (JT Baker ultrapure). We then added 0.4 mg of ammonium persulfate (Malinckrodt) 263 to each sample. We autoclaved the sample for 30 min at a pressure of 1.05 kg/cm2 and a 264 temperature of 122ºC. We measured total P by ICP-OES (Thermo iCAP6300) at a 265 wavelength of 213.6 nm. 266 267 3. Results & Discussion 268 3.1. Fate of P through lipid extraction 12 269 Freeze dried Synechocystis sp. PCC 6803 biomass contained 1.39%±0.28% total P 270 by dry mass. (All weights given by dry weight. ± indicates half standard deviation.) This 271 is consistent with previous findings that P is 1.5% of dry cell mass (Kim et al. 2010). In 272 lipid-extracted biomass samples, primary residual contained 1.50%±0.36% total P by dry 273 mass. Figure 2 summarizes the fate of P through lipid extraction normalized to 100 mg of 274 total P in the starting biomass. The primary residual contained 92±4.3 mg total P. Crude 275 lipid contained 7.3±4.2 mg total P. For transesterified samples, total P in the FAME was 276 0.5±0.1 mg total P. Total P in the secondary residual was 9.5±5.3 mg. Thus, nearly all of 277 the starting organic P was in the primary residual after lipid extraction. Of the small 278 amount in the crude lipids, nearly all of it was in the secondary residual. Essentially no P 279 (<1% of the starting P) was in the transesterified FAME. 280 These findings support our expectation that nucleic acid is the primary storage of 281 total cell P, with only small amounts stored in phospholipids. P associated with 282 phospholipid partitions to the crude lipid during extraction, while P associated with nucleic 283 acid remains in the primary residual. This explains the large fraction of P found 284 experimentally in the primary residual. The observed increase in P content from dry cells 285 to primary residual (1.39±0.28% to 1.50±0.36%) was not statistically significant, but any 286 increase would demonstrate the disproportional storage of P in non-lipid structures. The 287 92±4% of P found experimentally in the residual correlates with the expected 98% P 288 associated with nucleic acid. We attribute the small amount of P found in the fatty acids to 289 impurities from incomplete partitioning and analytical margin of error. 290 291 3.2. Oxidation of Organic P to Release Ortho-PO43- 13 292 Since only small amounts of the starting P were in the crude lipid and subsequent 293 lipid processing, the primary residual became the focus for P recovery. Prior to treatment 294 with H2O2 and microwave heating, this primary residual contained 82±1 mg total P with 295 0.2 mg of it as ortho-PO43-. After H2O2 and microwave treatment, samples contained 296 90±12 mg total P, including 75±6 mg as ortho-PO43-. Therefore, H2O2 oxidation recovered 297 106±17% of the total P (analytical error accounts for recovery over 100%) and released 298 most of it as ortho-PO43-, which was the objective. 299 300 301 3.3. Recovery of Ortho-PO43- by Resins from DI Water Figure 3A shows the ability of the two resins to absorb P in DI water. Both resins 302 were able to capture nearly all of the influent P up to 30 bed volumes. At this point, the 303 capacity of the resins was 5.0 mgP/g resin and 4.7 mgP/g resin for the HAX and SBAX 304 resins, respectively. The HAX resin then began a sharp breakthrough and reached 305 complete saturation near 80 bed volumes. The SBAX resin began a gradual breakthrough, 306 reaching 50% saturation around 200 bed volumes and 80% saturation around 500 bed 307 volumes. At the end of the experiments, the HAX resin sorbed a total mass of 38 mg of P, 308 giving a sorption capacity of 11 mgP/g resin, and the SBAX resin sorbed a total mass of 309 140 mg of P, giving a sorption capacity of 40 mgP/g resin. 310 Both resins released all of the P that would be eluted within the first 20 bed 311 volumes of regeneration. They did not release any additional P with 10 additional bed 312 volumes of regeneration (Figure 3B). The fastest rate of P elution for the SBAX resin 313 occurred around 5 bed volumes, and around 8 bed volumes for the HAX resin. A total of 314 19 mg of P was eluted from the HAX resin, or 51% of the total sorbed P was recovered. A 14 315 total of 167 mg of P was eluted from the SBAX resin, or 119% of the total sorbed was 316 recovered (the lack of mass-balance closure was due to analytical error from high dilution 317 required for analysis of concentrated elute). The pH of the HAX elute containing the 318 recovered P was 12, and of the SBAX elute it was 6. 319 The HAX resin had higher selectivity for P as demonstrated by the lower amount of 320 P in the column effluent, the sharp breakthrough curve showing a short saturation zone, 321 and the higher sorption capacity. We therefore expect it to have a higher rate of P capture 322 in solutions with competing constituents like the oxidized biomass. However, 0.1 N KOH 323 did not efficiently recover the sorbed P. While the iron nanoparticles lead to higher 324 sorption capacity than SBAX, they apparently made it more difficult to desorb the P. Poor 325 recovery might indicate that at least part of the sorbed P was irreversibly adsorbed by the 326 impregnated iron (hydr)oxide nanoparticles instead of sorbed entirely by anion exchange. 327 Our result differs from previous studies that showed that 80-90% of the P could be released 328 by elution from the HAX resin (Martin et al. 2009, Sengupta 2013) using 0.5-1.0 N NaOH 329 plus 0.4 N NaCl. Differences with these previous studies include different influent 330 matrices, not using combined NaCl and NaOH elutes or in as strong doses, and lower resin 331 contact time. We avoided stronger eluent doses so the recaptured P would not be in such a 332 high saline or high pH matrix that it would be unsuitable for subsequent microbial growth. 333 Since elution of the SBAX resin with 0.1 N NaCl showed the best recovery, we focused 334 our subsequent ion-exchange work on it. 335 In order to improve performance with the SBAX resin, we varied column operation 336 parameters to improve the P capture and release. For P capture, a steep breakthrough 337 curve is desired so that all of the P is captured until the inception of breakthrough, at which 15 338 time the column is stopped and regenerated. The SBAX breakthrough curve could be 339 made steeper by lowering the hydraulic loading rate. Figure 4 shows results for a SBAX 340 column receiving 100 mgP/L influent in DI water with an EBCT of 20 min (instead of 2 341 min) and a lower hydraulic loading rate of 3 BV/hr (instead of 30 BV/hr). Consequently, 342 the resin captured all ortho-PO43- for 200 BV before exhibiting a steep and desirable 343 breakthrough curve. This gave a sorption capacity of 35.6 mgP/g resin. For P 344 regeneration, slower elution (2 BV/hr) gave 99% recovery of the loaded P within 4 BVs. 345 This allowed us to achieve an 80-fold increase in P concentration in the regenerant. 346 Additional tests (data not shown) indicated greater ortho-PO43- exchange capacity at pH 5 347 instead of 8. This effort aimed to show that each step in this proof-of-concept P-recovery 348 sequence could be optimized to obtain desired performance outcomes. 349 350 351 3.4. Recovery of Ortho-PO43- by Resins from Oxidized Biomass We pumped oxidized primary residual through the ion exchange columns with 352 enough resin so the influent did not exceed 20 bed volumes to ensure complete capture of 353 the P. The HAX column effluent contained 1.7±0.3 mg of P out of the 72±0.9 mgP 354 influent, indicating 98% P capture on the resin. After elution, 16.7±0.0 mg P was in the 355 100 mL elute. Of this, 14.9±0.1 mg was ortho-PO43-. The pH of the pooled elute was 356 12.4±0.5. Overall, the HAX resin recovered 23%±0.2% of the influent P to the 357 regeneration solution. 358 The SBAX column effluent contained 20.9±7.6 mg of P out of 108±7.6 mgP 359 influent, indicating 81% of the P sorbed to the resin. After elution, 54.4±8.9 mg of P was 360 in the 100 mL elute. Of this, 53.0±8.2 mg was ortho-PO43-. The pH of the pooled elute 16 361 was 6.6±0.1. Overall, the SBAX resin recovered 50%±5% of the influent P to the 362 regeneration solution. 363 Both resins were only able to recover about half as much P when loaded from 364 oxidized biomass as opposed to when loaded from DI water: HAX went from 51% to 365 23%, and SBAX went from 119% to 50%. Previous studies have also observed lower 366 recovery from complex solutions like sludge liquor than from synthetic solutions (Bottini 367 and Rizzo 2012). In addition to ortho-PO43-, the solutions from the oxidized biomass also 368 contained residual organic matter (after oxidation 15 mg P out of 90 mg P was still 369 organic-bound) and other anions (bicarbonate, carbonate, sulfate, and nitrate) that were 370 probably also exchanged by the resins. Additionally, the influent pH for DI tests was 5, 371 but for influent oxidized biomass it was over 6. Having the pH approach the second 372 deprotonation for ortho-PO43- (pKa,2 = 7.2) during loading shifted a small fraction of its 373 speciation away from the single charge H2PO4- to the double charged HPO4-2. This may 374 have reduced ortho-PO43- adsorption capacity because each HPO4-2 takes up two anion- 375 exchange sites. This effect would be even stronger during regeneration due to the higher 376 pH (12 for the HAX) of the elute when almost all of the ortho-PO43- would be present as 377 HPO4-2. In the case of the HAX resin, this competition for anion exchange sites may have 378 forced more ortho-PO43- to be sorbed to the iron (hydr)oxide nanoparticles which could 379 form inner sphere complexes with stronger bonding and less elution. 380 381 382 383 3.5. P Recovery and Reuse Figure 5 summarizes results for each process step in the overall recovery process using the SBAX resin. The lipid extraction, cellular oxidation, and nutrient isolation steps 17 384 were, respectively, able to recover 93%, 106%, and 50% (using SBAX) of the starting P. 385 The overall process recovered 54% of the starting intracellular P into a pure and 386 concentrated nutrient solution. This yield is similar to other systems designed for complete 387 P recovery (Blocher et al. 2012) and shows that nutrient reuse in the context of microalgae 388 biofuel production is viable. 389 The recovered solution had an ortho-PO43- concentration of 10.6 mgP/L, compared 390 to 5.4 mgP/L required in standard BG-11. We also measured 0.95 mg NO3 --N/L and 1.5 391 mg SO42--S/L, compared to 247 and 9.8 mg/L required for BG-11, respectively, 392 demonstrating the selectivity of the resin for P. 393 The P solution recovered from the SBAX supported cyanobacteria growth. The 394 optical density increased from 0.12 initially to 0.55 after one day and to 1.11 after one 395 week. This correlates to specific growth rates of 1.4 day-1 over one day and 0.7 day-1 over 396 one week. For comparison, the optical density of the same cell culture grown in a BG-11 397 solution without any P went from 0.12 initially to 0.16 after one day and 0.10 after one 398 week, corresponding to specific growth rates of 0.26 day-1 after one day and -0.06 day-1 399 after one week. The nearly ten-fold increase in cell density over one week in the solution 400 containing recovered P confirms that the recovered P was available for cyanobacteria 401 uptake. It also demonstrates that we did not co-recover any substances that would inhibit 402 reuse, such as harmful heavy metals or residual oxidant. These rates are comparable to 403 growth rates previously observed for Synechocystis using BG-11 (Kim et al. 2010) albeit in 404 a different reactor configuration. 405 406 We recommend future work improving P release methods that can co-recover other valuable products produced by cyanobacteria, like other nutrients, proteins, or ethanol 18 407 (Wijffels et al. 2013). We further recommend improving P capture efficiency, reducing the 408 overall cost, energy, and chemical footprint of the process, and demonstrating recovery on 409 full-scale. Other future work could compare the effectiveness of growing microalgae on 410 recovered P compared to other sources of P with complete controls. 411 412 4. Conclusions Efficient P recycling in microbial biofuel production will be essential to preventing 413 414 415 competition between food and energy systems. This work demonstrates: • is in nucleic acids, with very little in phospholipids. 416 417 • • 422 While HAX resin showed higher affinity for ortho-PO43-, the SBAX resin released the ortho-PO43- more completely. 420 421 Advanced oxidation transformed over 80% of that organic P into useful and recoverable ortho-PO43-. 418 419 After lipid processing, over 90% of the P remained in the residuals. Most cellular P • Both resins recovered less P from oxidized biomass than from P spiked DI water, likely due to interference with residual organics or competing oxyanions. 423 424 425 Acknowledgements A Dean’s Fellowship from the Ira A. Fulton Schools of Engineering at Arizona 426 State University provided partial funding for this study, as did the Central Arizona Phoenix 427 Long Term Ecological Research (CAP LTER) project from the National Science 428 Foundation (BCS-1026865). Thank you to Jie Sheng who provided training on lipid 19 429 extraction and biofuel processing. Thank you to Chao Zhou and Levi Straka for culturing 430 the cyanobacteria. 431 20 432 References 433 Batan, L., Quinn, J., Willson, B. and Bradley, T. (2010) Net Energy and Greenhouse Gas 434 Emission Evaluation of Biodiesel Derived from Microalgae. Environmental 435 Science and Technology 44(20), 7975-7980. 436 Blaney, L.M., Cinar, S. and Sengupta, A.K. (2007) Hybrid Anion Exchanger for Trace 437 Phosphate Removal from Water and Wastewater. Water Research 41(7), 1603- 438 1613. 439 Blocher, C., Niewersch, C. and Melin, T. (2012) Phosphorus Recovery from Sewage 440 Sludge with a Hybrid Process of Low Pressure Wet Oxidation and Nanofiltration. 441 Water Research 46(6), 2009-2019. 442 Bottini, A. and Rizzo, L. (2012) Phosphorus Recovery from Urban Wastewater Treatment 443 Plant Sludge Liquor by Ion Exchange. Separation Science & Technology 47(4), 444 613-620. 445 Childers, D.L., Corman, J., Edwards, M. and Elser, J.J. (2011) Sustainability Challenges of 446 Phosphorus and Food: Solutions from Closing the Human Phosphorus Cycle. 447 Bioscience 61(2), 117-124. 448 Clarens, A.F., Resurreccion, E.P., White, M.A. and Colosi, L.M. (2010) Environmental 449 Life Cycle Comparison of Algae to Other Bioenergy Feedstocks. Environmental 450 Science & Technology 44(5), 1813-1819. 451 452 Cordell, D., Drangert, J. and White, S. (2009) The Story of Phosphorus: Global Food Security and Food for Thought. Global Environmental Change 19(2), 292-305. 21 453 de-Bashan, L.E. and Bashan, Y. (2004) Recent advances in removing phosphorus from 454 wastewater and its future use as fertilizer (1997-2003). Water Research 38(19), 455 4222-4246. 456 Elser, J.J., Acharya, K., Kyle, M., Cotner, J., Makino, W., Markow, T., Watts, T., Hobbie, 457 S., Fagan, W., Schade, J., Hood, J. and Sterner, R.W. (2003) Growth Rate 458 Stoichiometry Couplings in Diverse Biota. Ecology Letters 6(10), 936-943. 459 Erisman, J.W., van Grinsven, H., Leip, A., Mosier, A. and Bleeker, A. (2010) Nitrogen and 460 biofuels; an overview of the current state of knowledge. Nutrient Cycling in 461 Agroecosystems 86(2), 211-223. 462 Folch, J., Lees, M. and Stanley, G.H.S. (1957) A Simple Method for the Isolation and 463 Purification of Total Lipids from Animal Tissues. Journal of Biological Chemistry 464 226(1), 497-509. 465 466 467 468 469 470 471 Geider, R. and La Roche, J. (2002) Redfield revisited: Variability of C:N:P in marine microalgae and its biochemical basis. European Journal of Phycology 37(1), 1-17. Hajime, W. and Murata, N. (2007) The Essential Role of Phosphatidylglycerol in Photosynthesis. Photosynthesis Research 92(2), 205-215. Huo, Y.-X., Wernick, D.G. and Liao, J.C. (2012) Toward nitrogen neutral biofuel production. Current Opinion in Biotechnology 23(3), 406-413. Kaneko, T., Sato, S., Kotani, H., Tanaka, A., Asamizu, E., Nakamura, Y., Miyajima, N., 472 Hirosawa, M., Sugiura, M., Sasamoto, S., Kimura, T., Hosouchi, T., Matsuno, A., 473 Muraki, A., Nakazaki, N., Naruo, K., Okumura, S., Shimpo, S., Takeuchi, C., 474 Wada, T., Watanabe, A., Yamada, M., Yasuda, M. and Tabata, S. (1996) Sequence 475 Analysis of the Genome of the Unicellular Cyanobacterium Synechocystis sp. 22 476 Strain PCC6803. II. Sequence Determination of the Entire Genome and 477 Assignment of Potential Protein-coding Regions. DNA Research 3(3), 109-136. 478 Kim, W.K., Vannela, R., Zhou, C., Harto, C. and Rittmann, B.E. (2010) Photoautotrophic 479 Nutrient Utilization and Limitation During Semi-Continuous Growth of 480 Synechocystis sp. PCC6803. Biotechnology and Bioengineering 106(4), 553-563. 481 Liao, P.H., Wong, W.T. and Lo, K.V. (2005) Advanced Oxidation Process Using 482 Hydrogen Peroxide/Microwave System for Solubilization of Phosphate. Journal of 483 Environmental Science and Health 40(9), 1753-1761. 484 Martin, B.D., Parsons, S.A. and Jefferson, B. (2009) Removal and Recovery of Phosphate 485 from Municipal Wastewater Using a Polymeric Anion Exchanger Bound with 486 Hydrated Ferric Oxide Nanoparticles. Water Science and Technology 60(10), 487 2637-2645. 488 Mata, T.M., Martins, A.A. and Caetano, N.S. (2010) Microalgae for Biodiesel Production 489 and Other Applications: A Review. Renewable & Sustainable Energy Reviews 490 14(1), 217-232. 491 Midorikawa, I., Aoki, H., Omori, A., Shimizu, T., Kawaguchi, Y., Kassai, K. and 492 Murakami, T. (2008) Recovery of High Purity Phosphorus from Municipal 493 Wastewater Secondary Effluent by a High Speed Adsorbent. Water Science and 494 Technology 58(8), 1601-1607. 495 496 Miner, G. (2006) Standard Methods for the Examination of Water and Wastewater, 21st Edition, American Water Works Association. 497 Morse, G.K., Brett, S.W., Guy, J.A. and Lester, J.N. (1998) Review: Phosphorus Removal 498 and Recovery Technologies. The Science of the Total Environment 212(1), 69-81. 23 499 500 Rees-Nowak, D., Marston, C. and Gisch, D. (2005) Controlling Chromium. Water & Wastewater International 20(5), 21. 501 Rippka, R., Deruelles, J., Waterbury, J.B., Herdman, M. and Stanier, R.Y. (1979) Generic 502 Assignments, Strain Histories And Properties Of Pure Cultures Of Cyanobacteria. 503 Journal of General Microbiology 111(Mar), 1-61. 504 505 506 507 508 Rittmann, B.E. (2008) Opportunities for renewable bioenergy using microorganisms. Biotechnology and Bioengineering 100(2), 203-212. Rittmann, B.E., Mayer, B., Westerhoff, P. and Edwards, M. (2011) Capturing the lost phosphorus. Chemosphere 84(6), 846-853. Sakurai, I., Shen, J., Leng, J., Ohashi, S., Kobayashi, M. and Wada, H. (2006) Lipids in 509 Oxygen Evolving Photsystems II Complexes of Cyanobacteria and Higher Plants. 510 Journal of Biochemistry 140(2), 201-209. 511 Schenk, P.M., Thomas-Hall, S.R., Stevens, E., Marx, U.C., Mussgnug, J.H., Posten, C., 512 Kruse, O. and Hankamer, B. (2008) Second Generation Biofuels: High Efficiency 513 Microalgae for Biodiesel Production. Bioenergy Research 1(1), 20-43. 514 Sengupta, S. (2013) Novel Solutions to Water Pollution. Ahuja, S. and Hristovski, K. 515 (eds), pp. 167-187, American Chemical Society, Washington DC. 516 Shastri, A.A. and Morgan, J.A. (2005) Flux Balance Analysis of Photoautotrophic 517 518 Metabolism. Biotechnology 21(6), 1617-1626. Sheng, J., Kim, H.W., Badalamenti, J.P., Zhou, C., Sridharakrishnan, S., Krajmalnik- 519 Brown, R., Rittmann, B.E. and Vannela, R. (2011a) Effects of Temperature Shifts 520 on Growth Rate and Lipid Characteristics of Synechocystis sp. PCC6803 in a 521 Bench-Top Photobioreactor. Bioresource Technology 102(24), 11218-11225. 24 522 Sheng, J., Vannela, R. and Rittmann, B.E. (2011b) Evaluation of Methods to Extract and 523 Quantify Lipids from Synechocystis PCC6803. Bioresource Technology 102(2), 524 1697-1703. 525 526 527 Soh, L. and Zimmerman, J. (2011) Biodiesel Production: The Potential of Algal Lipids Extracted with Supercritical Carbon Dioxide. Green Chemistry 13(6), 1422-1429. Sterner, R.W. and Elser, J.J. (2002) Ecological Stoichiometry: The Biology of Elements 528 from Molecules to the Biosphere, Princeton University Press, Princeton, New 529 Jersey. 530 Stucker, V., Ranville, J., Newman, M., Peacock, A., Cho, J. and Hatfield, K. (2011) 531 Evaluation and application of anion exchange resins to measure groundwater 532 uranium flux at a former uranium mill site. Water Research 45(16), 4866-4876. 533 534 Takahashi, H., Uchimiya, H. and Hihara, Y. (2008) Difference in metabolite levels 535 between photoautotrophic and photomixotrophic cultures of Synechocystis sp. PCC 536 6803 examined by capillary electrophoresis electrospray ionization mass 537 spectrometry. Journal of Experimental Botany 59(11), 3009-3018. 538 United States Environmental Protection Agency (USEPA) (2008) Test Methods for 539 Evaluating Solid Waste, Physical/Chemical Methods, Washington D.C. 540 United States Geological Survey (USGS) (2011) Mineral Commodity Summaries 2011, 541 Reston, VA. 542 van de Meene, A.M.L., Hohmann-Marriott, M.F., Vermaas, W.F.J. and Roberson, R.W. 543 (2006) The three-dimensional structure of the cyanobacterium Synechocystis sp 544 PCC 6803. Archives of Microbiology 184(5), 259-270. 25 545 Vermaas, W. (1996) Molecular Genetics of the Cyanobacterium Synechocystis sp. PCC 546 6803: Principles and Possible Biotechnology Applications. Journal of Applied 547 Phycology 8(4-5), 263-273. 548 549 550 Vermaas, W.F.J. (2001) Encyclopedia of Life Sciences, pp. 245-251, John Wiley & Sons, Ltd, London. Wijffels, R.H., Kruse, O. and Hellingwerf, K.J. (2013) Potential of Industrial 551 Biotechnology with Cyanobacteria and Eukaryotic Microalgae. Current Opinion in 552 Biotechnology 24(3), 405-413. 553 Wong, W.T., Chan, W.I., Liao, P.H., Lo, K.V. and Mavinic, D.C. (2006) Exploring the 554 Role of Hydrogen Peroxide in the Microwave Advanced Oxidation Process; 555 Solubilization of Ammonia and Phosphates. Journal of Environmental Engineering 556 and Science 5(6), 459-465. 557 26 O 27% Elemental H 7% C 49% Biological Carbohydrate 19% Protein 51% Phosphorus N 12% P 1.5% S 1% Ash 3% Lipid 10% Lipids 3% DNA 15% RNA 83% DNA 3% RNA 17% Figure 1 – The estimated location of P within Synechocystis sp. PCC 6803 shown on the right determined by the elemental (Kim et al. 2010) and biological (Shastri and Morgan 2005) composition shown on the left. All numbers given are percent by weight of the total biomass (left) or total P in the biomass (right). A majority of cellular P is in RNA, and only small amounts are in lipids. Thus, almost all P is in the primary residuals after lipid extraction, not in the lipid extract. Biomass 100mgP Supernatant 0mgP Primary Residual 92±4mgP Sample Prep Biomass Pellet 100mgP Extraction Crude Lipid Secondary Residual 9±5mgP 7±4mgP Transesterification Fatty Acid Methyl Ester 1±0mgP Figure 2 – The fate of 100 mg of starting P through the lipid extraction process. Most of the P remained with the biomass in the primary residual, although some was associated with the crude lipid remains in the secondary residual. The FAME only contained about 1% of the starting P. 3A 80% HAX SBAX 60% 30 BV Effluent Concentration per Influent Concentration 100% 40% 20% 0% 0 50 100 Bed Volumes Treated 150 200 Cumulative Fraction Desorbed 120% 100% 3B SBAX HAX 80% 60% 40% 20% 0% 0 5 10 Bed Volumes Treated 15 20 Figure 3 – Performance of an iron hydr(oxide) impregnated anion exchange (HAX) resin (squares) and a strong-base anion exchange (SBAX) resin (diamonds) for recovering phosphate from DI water. (A) Uptake of phosphate by fresh resin in column test. Uses hydraulic loading rate of 30 BV/hr, an initial P concentration of 80 mgP/L, and influent pH 5. (B) Desorption of phosphate from resin by 0.1 N KOH for HAX or 0.1 N NaCl for SBAX with hydraulic loading rate of 6 BV/hr, normalized to mass of P sorbed. The HAX resin shows higher affinity for P during sorption, but the SBAX releases more P upon elution. Effluent Concentration per Influent Concentration 100% 4A 80% 60% 40% 20% 0% Cumulative Fraction Desorbed 0 100% 100 200 300 Bed Volumes Treated 4B 80% 60% 40% 20% 0% 0 2 4 Bed Volumes Treated 6 Figure 4 – Enhanced P recovery from DI water on SBAX resin by improving operating conditions. (A) Uptake of phosphate by fresh resin in column test. Uses hydraulic loading rate of 3 BV/hr, an initial P concentration of 100 mgP/L, and influent pH 8. (B) Desorption of phosphate from resin by 1 N NaCl at a hydraulic loading rate of 2 BV/hr, normalized to mass P sorbed. The steep breakthrough after a long bed run is optimal for P recovery, and subsequent elution in few bed volumes gives an 80-fold increase in P concentration. Starting Biomass 100 mg total P 0 mg ortho-PO4-3 Primary Residual 93±2 mg total P 0 mg ortho-PO4-3 Oxidized Biomass 90±12 mg total P 75±6 mg ortho-PO4-3 Recovered Nutrients 54±9 mg total P 53±8 mg ortho-PO4-3 Figure 5 – Process step yields of total P and ortho-PO43- for 100 mg starting P through the P-recovery process using advanced oxidation and SBAX. Nearly all cellular P was found in the primary residual after lipid extraction. Advanced oxidation transformed a majority of the P to recoverable and beneficial ortho-PO43-. SBAX resin could then sorb and elute a concentrated nutrient solution. The overall tested P-recovery process could capture more than 50% of the starting P in a beneficial form. Phosphorus Recovery Process